Structure of the Activated Edc1-Dcp1-Dcp2-Edc3 Mrna Decapping Complex with Substrate Analog Poised for Catalysis

ARTICLE

DOI: 10.1038/s41467-018-03536-x OPEN Structure of the activated Edc1-Dcp1-Dcp2-Edc3 mRNA decapping complex with substrate analog poised for catalysis

Jeffrey S. Mugridge1, Ryan W. Tibble1,2, Marcin Ziemniak3,4,5, Jacek Jemielity4 & John D. Gross 1,2

The conserved decapping enzyme Dcp2 recognizes and removes the 5′ eukaryotic cap from mRNA transcripts in a critical step of many cellular RNA decay pathways. Dcp2 is a dynamic 1234567890():,; enzyme that functions in concert with the essential activator Dcp1 and a diverse set of coactivators to selectively and efﬁciently decap target mRNAs in the cell. Here we present a 2.84 Å crystal structure of K. lactis Dcp1–Dcp2 in complex with coactivators Edc1 and Edc3, and with substrate analog bound to the Dcp2 active site. Our structure shows how Dcp2 recognizes cap substrate in the catalytically active conformation of the enzyme, and how coactivator Edc1 forms a three-way interface that bridges the domains of Dcp2 to consolidate the active conformation. Kinetic data reveal Dcp2 has selectivity for the ﬁrst transcribed nucleotide during the catalytic step. The heterotetrameric Edc1–Dcp1–Dcp2–Edc3 structure shows how coactivators Edc1 and Edc3 can act simultaneously to activate decapping catalysis.

1 Department of Pharmaceutical Chemistry, University of California, San Francisco, San Francisco, CA 94158, USA. 2 Program in Chemistry and Chemical Biology, University of California, San Francisco, San Francisco, CA 94158, USA. 3 Division of Biophysics, Institute of Experimental Physics, Faculty of Physics, University of Warsaw, 02-089 Warsaw, Poland. 4 Centre of New Technologies, University of Warsaw, 02-097 Warsaw, Poland. 5Present address: Biological and Chemical Research Centre, Department of Chemistry, University of Warsaw, 02-089 Warsaw, Poland. Correspondence and requests for materials should be addressed to J.D.G. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x

egradation of mRNA transcripts plays a critical role in Dcp2 NRD as in our Mugridge et al. substrate-bound structure, maintenance of the cellular transcriptome and in control but the CD was rotated by ~90° to bring the Nudix helix much D 1 7 7 of eukaryotic gene expression . Removal of the m GpppN closer to the cleaved m GDP product, consistent with an active cap structure found at the 5′ end of all eukaryotic mRNA marks conformation of Dcp222. Finally, most recently Wurm et al. these transcripts for rapid clearance via the 5′-to-3′ mRNA decay published a study that presented an S. pombe Edc1–Dcp1–Dcp2 pathway2. The conserved mRNA-decapping enzyme Dcp2 cata- product-bound structure in the same conformation as observed lyzes hydrolysis of the 5′ cap to liberate m7GDP and 5′ mono- by Charenton et al., a conformation that is significantly populated phosphate RNA, which can then be degraded by conserved 5′-3′ in solution when Edc1 and m7GDP or capped RNA are bound19. – exonucleases3 6. Dcp2-mediated decapping is a critical step in In this study, we present a new, four-protein structure of K. many cellular RNA decay pathways including bulk 5′–3′ decay4,5, lactis Edc1–Dcp1–Dcp2–Edc3 at 2.84 Å resolution in the cataly- – nonsense-mediated decay7 10, microRNA-mediated decay11,12, tically active conformation with bound substrate analog poised decay of long-noncoding RNA13, and decay of transcripts con- for catalysis. We show that the conserved YAGxxF activation taining non-optimal codons14. Recently, Dcp2 was also shown to motif of Edc1 mediates a three-way interface between Dcp1 and be a reader of reversible m6A modifications in the 5′ cap, whereby the Dcp2 NRD and CD, and that lesions on the NRD and in the m6A methylation inhibits decapping and stabilizes cellular flexible hinge linking the CD and NRD abolish activation by mRNA transcripts15. Edc1. Substrate analog binding to the active conformation of Dcp2 consists of an N-terminal regulatory domain (NRD) and Dcp2 suggests that the first transcribed nucleotide of RNA is a catalytic Nudix domain (CD)16,17 linked by a short, flexible recognized by base stacking to a conserved aromatic residue at the hinge that allows the enzyme to dynamically adopt different top of the RNA-binding channel, and that the RNA body follows – conformations in solution18,19 and in crystal structures20 24. The this positively charged path to bound Edc3 coactivator. We show CD harbors a positively charged patch for RNA binding (box B that Dcp2 has selectivity for the first transcribed nucleotide of motif)25,26 and a Nudix motif with conserved Glu residues that RNA substrate and that this selectivity depends on the conserved bind magnesium and carry out cap hydrolysis chemistry27,28. aromatic residue that binds this nucleotide of our substrate Fungal Dcp2 also contains an unstructured C-terminal extension analog. Overall, these data show how substrate binds and is that harbors protein–protein interaction motifs and elements that recognized by the catalytically active conformation of Dcp2, inhibit Dcp2 catalysis29. At least some of these motifs are trans- suggest a path for RNA binding consistent with prior in vitro ferred to the C-terminus of the essential activator Dcp1 in mutagenesis and binding data, and show how coactivators Edc1 metazoans, suggesting a conserved biological function30. The and Edc3 can simultaneously bind Dcp2 to enhance mRNA NRD enhances decapping activity by specifically recognizing the decapping by the Dcp1–Dcp2 complex. Our structure establishes m7G nucleotide of cap and binding Dcp1, which further accel- that the active conformation of Dcp2 is the same for both sub- erates decapping17,19,21,22,31,32. strate and product-bound complexes, and clarifies the mechan- The activity of the core Dcp1–Dcp2 complex can be further isms of decapping catalysis and control of decapping activation by modulated by the enhancer of decapping proteins (Edcs)33. For multiple protein interactions. example, the Edc1-type activators, which include the yeast paralog Edc2 and PNRC2 in metazoans, contain conserved tandem short linear motifs to bind Dcp1 and activate Dcp2 cata- Results – – lysis19 21,34 37. Edc3 activates decapping by promoting RNA Structure of substrate analog-bound Edc1–Dcp1–Dcp2–Edc3. – binding and alleviating autoinhibition of Dcp2 in yeast22,29,38 42. To obtain our recently published substrate analog-bound Edc3 is recruited to the decapping complex via short linear motifs PNRC2–Dcp1–Dcp2 structure21, we crystallized an apo found in the fungal or metazoan Dcp2 or Dcp1 C-terminal PNRC2–Dcp1–Dcp2 complex and soaked in cap analog to induce extension, respectively29,33. Edc4 is a metazoan-specific scaf- a conformational change and substrate binding within the crystal. folding protein that strengthens Dcp1–Dcp2 interaction and While this substrate-bound conformation gave new insight into coordinates decapping with exonucleolytic degradation by cap recognition by Dcp2, it was clear that this did not represent – Xrn132,43 45. Genetic and physical interactions between activator the catalytically active conformation because the catalytic Glu proteins suggest these cofactors work together to promote dec- residue on the Nudix helix was positioned to too far from bound apping, but how this is achieved at the structural level is substrate to carry out hydrolysis chemistry. When we attempted – unknown45 49. to push Dcp2 into the active conformation by soaking in cata- Several recent structural studies have provided new insights lytically required magnesium along with substrate analog, the into how Dcp2 recognizes the m7G cap, how Dcp2 activity is crystals were physically damaged and diffraction was completely enhanced by coactivators, and how Dcp2 solution conformations destroyed, suggesting this induced an additional conformational are tied to catalysis. First, a study by Valkov et al. presented a change in Dcp2 that could not be accommodated by the crystal structure of S. pombe Edc1–Dcp1–Dcp2 in which the activation lattice. motif in Edc1 bridges the Dcp2 NRD and CD to apparently To trap the substrate-bound active conformation, we screened promote a conformation of Dcp2 that enhances RNA binding20. different Dcp1–Dcp2 constructs (S. pombe, K. lactis, H. sapiens) Soon after, our lab published the first substrate analog-bound in which the catalytic Glu was mutated to Gln to prevent cap structure of S. pombe Dcp1–Dcp2 in complex with human hydrolysis for co-crystallization with magnesium, a tight-binding PNRC221. Our structure showed a very different conformation two-headed cap analog50, and coactivators Edc1 and Edc3 than the Valkov structure, in which the substrate analog was (Fig. 1a,b). The K. lactis constructs crystallized most readily and bound by a composite binding site using conserved residues on we obtained a structure of Kl Edc1–Dcp1–Dcp2–Edc3 with the NRD and CD. However, in this pre-catalytic conformation of bound magnesium and two-headed cap analog at 2.84 Å Dcp2 the catalytic Glu on the Nudix helix was positioned too far resolution (Fig. 1c, Table 1, solved by molecular replacement from substrate for effective catalysis, and the activation motif of with PDB 5LOP22 Dcp1–Dcp2). Dcp2 is in the catalytically active PNRC2 was disordered, suggesting an additional conformational conformation, very similar to that found in recently reported change in Dcp2 was needed to achieve catalysis. Simultaneously, m7GDP product-bound structures (Supplementary Fig. 1; Dcp2 Charenton et al. published a K. lactis Dcp1–Dcp2–Edc3 structure core backbone RMSD is 1.2 Å with PDB 5LOP and 1.9 Å with with bound m7GDP product in which m7G was recognized by the PDB 5N2V)19,22, with the Nudix helix positioned close to bound

2 NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x ARTICLE

abDcp2 activating Dcp1 binding OH OH O + N NH Kl Edc1(355–380) HSKCYAGATFATEAPQV–TTLPKPSFV– – O O O O O Sc Edc1(148–175) LSHSYAGSTFATNGPREAKNLPKPSFL –– H N N O P O P O P O P O N N NH 2 N O 2 S– O– O– S– Sc Edc2(119–145) GHTAFAGASFTTDIPHE –AAL PKPSFV – – HN N+ Sp Edc1(153–179) NSILYAGPTFTHSP– – AASNLP IPTFLHS O OH OH Hs PNRC1(274–301) QKENYAGAKFSDPP– – SPSVLPKPPSHWM Hs PNRC2(88–115) ANQNYAGAKFSEPP– – SPSVLPKPPSHWV

c Dcp1 d e Dcp1 Edc1

Edc1 F364 F Dcp2 NRD

Y359 R39 130 s loop Edc3 G361 G

F42 Dcp2 CD A I135 Y A360 Dcp2 NRD E152Q Y98 I102 P103 Dcp2 CD

Fig. 1 Structure of the activated decapping complex poised for catalysis. a Structure of the high-affinity, two-headed cap substrate analog. b Sequence alignment of Edc1-like coactivator peptides showing the conserved Dcp2-activating and Dcp1-binding motifs; background color intensity is scaled to conservation. c Structure of the substrate analog-bound Kl Edc1–Dcp1–Dcp2–Edc3 complex. Edc1 is brown, Dcp1 is yellow, Dcp2 NRD is purple, Dcp2 CD is green, Edc3 Lsm domain is dark blue, substrate analog is cyan. d Close-up view of the three-way interface between Dcp1, Dcp2 NRD, and Dcp2 CD mediated by the Edc1 YAGxxF Dcp2-activating motif. Fo–Fc omit map for the Edc1 peptide is shown at 2.5σ. Sidechains for the Edc1 peptide and nearby conserved residues on Dcp2 are shown as sticks. The catalytic Glu on the Nudix helix, mutated in our structure to Gln, is labeled as E152Q in red. e Surface view of the Edc1 activation motif–Dcp1–Dcp2 interface, with conserved residues in the activation motif labeled cap. Edc3 Lsm domain binds an extended helix at the C-terminus fully activated, substrate analog-bound structure. Tyr359 of Kl of the Dcp2 CD, as in the recent Kl Dcp1–Dcp2–Edc3 structure22, Edc1, the residue most critical for activation, makes a cation–π and the fully ordered Edc1 activation peptide bridges Dcp1 and interaction with the strictly conserved Arg39 on the Dcp2 NRD Dcp2, similar to the recent Sp Edc1–Dcp1–Dcp2 structure19. The (Fig. 1d). Ala360 packs against conserved hydrophobic residues in co-occupancy of Edc1 and Edc3 coactivators in our structure the flexible hinge between Dcp2 NRD and CD, and Phe364 plugs highlights the idea that these decapping factors work together to into a pocket at the Dcp1–Dcp2 NRD interface consisting of promote decapping by complementary mechanisms42. positively charged and hydrophobic residues (Supplementary Fig. 3). These contacts anchor the Edc1 activation motif to Dcp1–Dcp2 and provide a complementary surface for the flexible Edc1 consolidates the active conformation of Dcp2. Edc1-like hinge and 130s loop on the Dcp2 CD to pack against Edc1 in the coactivator proteins contain tandem, conserved short linear catalytically active conformation of the enzyme. In this way, motifs for Dcp1 binding and Dcp2 activation (Fig.1b)20,21,35,37. protein interactions mediated by the Dcp2 activation motif of Several structures have now been published that show how the Edc1 accelerate kcat not by contacting cap, but instead by limiting proline-rich Dcp1-binding motif in Edc1-like peptides binds a torsional degrees of freedom in the hinge to favor the active – conserved aromatic cleft on Dcp119 21,36, and our structure dis- conformation of Dcp2, in which conserved residues on the CD plays a very similar interaction between Kl Edc1 and Kl Dcp1 as and NRD recognize and hydrolyze cap using the composite active that observed in Sp, Hs, and chimeric Sp–Hs structures (Supple- site. mentary Fig. 2). The conserved YAGxxF Dcp2-activating motif creates a 3-way interface that bridges Dcp1, Dcp2 NRD, and Dcp2 CD in the catalytically active conformation by threading Residues in Dcp2 NRD and hinge are critical for activation.In through a narrow channel between these domains (Fig. 1d,e). In order to validate the Edc1–Dcp2 interface observed in our total, the Edc1 peptide interaction with the Dcp1–Dcp2 complex structure and better understand how the Edc1 peptide interacts buries ~1300 Å2 of surface area, with the YAGxxF Dcp2-activing with Dcp2 to promote catalysis, we mutated conserved residues motif accounting for ~500 Å2 of this buried surface. The interface on the Dcp2 NRD, CD, and hinge (Dcp2 residues shown as sticks between the Dcp2 CD and Edc1 peptide covers only ~250 Å2 of in Fig. 1d) that contact YAG residues of the Edc1 peptide, and surface area, consistent with evidence that the peptide alone is asked how these mutations affect activation of Dcp2 catalysis by insufficient to drive a conformational change in solution19. Edc1. Using single-turnover in vitro kinetic decapping assays with We have shown that Edc1-like proteins accelerate decapping cap-radiolabeled RNA substrate and S. pombe Edc1–Dcp1–Dcp2 21,35 fi catalysis by both increasing kcat and decreasing KM. Kinetic complex, we nd that residues Arg33 (Kl Arg39) and Ile96 (Kl analysis of the Dcp2-activating YAGxxF motif in PNRC2 showed Ile102) are absolutely essential for Edc1 function and that that mutation of the Tyr residue completely abolishes kcat mutation of these conserved residues to Ala or Gly, respectively, activation of Dcp2 catalysis, while mutation of Ala and Phe completely blocks activation by Edc1 (Fig. 2). In our structure, residues significantly impairs activation21. We can now rationa- Arg33 (Kl Arg39) makes a cation–π interaction with Edc1 Tyr359 lize our previous biochemical observations in the context of the to anchor the activation motif to Dcp2 NRD, and Ile96 (Kl Ile102)

NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x

ﬁ a Table 1 Crystallographic data collection and re nement 1.0 statistics

GDP 0.8 Kl Edc1–Dcp1–Dcp2–Edc3 with substrate analog 7 (PDB 6AM0) 0.6 WT Data collection R33A 0.4 F36A Space group C2 Y92A Cell dimensions I96G 0.2 P97G a, b, c (Å) 174.26, 83.25, 104.30 I130G α β γ Fraction product m , , (°) 90.00, 93.62, 90.00 0.0 Y220A Resolution (Å) 46.0–2.84 (3.00–2.84)a 02413 Rmerge 0.058 (0.78) I/σI 9.8 (1.3) Time (min) CC1/2 99.8 (58.5) Completeness (%) 98.7 (99.1) b Redundancy 2.9 (2.8) Reﬁnement 100 Edc1

Resolution (Å) 46.0–2.84 Sp No. reﬂections 35,286 Rwork/Rfree 0.2293 / 0.2491 No. atoms 10 Protein 8610 Ligand 115b Water – Fold activation by B factors 1 Protein 104

b WT

Ligand 114 I96G F36A Y92A R33A P97G I130G Water – Y220A R.m.s. deviations Fig. 2 Mutations on both the Dcp2 NRD and flexible hinge prevent Bond lengths (Å) 0.003 activation by Edc1. a In vitro decapping timecourses for Sp Edc1–Dcp1–Dcp2 Bond angles (°) 0.665 WT and Dcp2 mutant complexes. R33A and I96G mutations completely Ramachandran plot statistics block activation of Dcp1–Dcp2 by Edc1 peptide. Concentration of No. favored 967 (95.6 %) Dcp1–Dcp2(1–243) is 2 µM and Edc1(155–180) peptide is 100 µM. Errors No. allowed 45 (4.4 %) are s.d. of two independent replicates. b Fold activation of WT and Dcp2 No. outliers 0 (0 %) mutant decapping complexes by Edc1 peptide. Calculated as k (Edc1–Dcp1–Dcp2)/k (Dcp1–Dcp2). k values for Dcp1–Dcp2 with Data set was collected from a single crystal obs obs obs a Values in parentheses are for highest-resolution shell no Edc1 coactivator peptide are shown in Supplementary Fig. 4. Errors are s. b Two molecules of two-headed cap analog and one magnesium ion d. of kobs(Edc1–Dcp1–Dcp2)/kobs(Dcp1–Dcp2) values obtained from two independent replicates packs against Edc1 Ala360 and the Tyr359 backbone. These data suggest that both positioning of the peptide activation motif along Substrate analog binds a composite active site on Dcp2. Two- the Dcp1–Dcp2 NRD surface, and specific van der Waals inter- headed cap substrate analog bound at the Dcp2 active site engages actions with the Dcp2 hinge that links the NRD and CD, are conserved residues on the Dcp2 NRD and CD (Fig. 3a,b). In critically important for Edc1-mediated activation. agreement with our previously reported substrate analog-bound 21 7 Interestingly, mutation of other strictly conserved residues on pre-catalytic structure , and recently reported m GDP product- 19,22 7 Dcp2 that contact the Edc1 peptide in the catalytically active bound structures , the m G of cap is recognized by conserved, conformation, such as Phe36 (Kl Phe42) and Tyr92 (Kl Tyr98), catalytically critical Trp and Asp residues on the Dcp2 NRD. The 7 have very little effect on Edc1’s ability to activate Dcp2. Mutation opposite face of this m G nucleotide packs against the backbone of Tyr220, a distal residue on the CD that is important for of the Dcp2 CD 190s loop in the catalytically active conformation. conformational dynamics in Dcp242, also has no effect on This agrees with recent product-bound structures, but differs 7 activation by Edc1. Importantly, none of the tested mutations to from the pre-catalytic conformation, in which this face of m G conserved Dcp2 residues resulted in large defects in decapping in was engaged in base stacking with Sp Tyr220 (Kl Phe223). In the the absence of Edc1 (Supplementary Fig. 4), showing that Arg33 active conformation reported here however, the CD is rotated (Kl Arg39) and Ile96 (Kl Ile102) are critical specifically for Edc1- ~90° from the pre-catalytic conformation, and Phe223 base stacks 7 mediated activation of Dcp2. Finally, in contrast to our previous with the second m G of our two-headed substrate analog, likely fi observations examining Dcp2 activation in the chimeric Hs mimicking binding of the rst transcribed nucleotide of RNA PNRC2–Sp Dcp1–Dcp2 system where we showed PNRC2 substrate to the Dcp2 CD. Additional conserved and catalytically peptides with the YAGxxF Dcp2-activating motif had little important residues make contacts with the substrate analog phosphate chain, including the general acid Lys134 and Arg132 impact on the KM of the decapping complex, with native Sp – – on the CD, and Lys99 on the NRD. Edc1 Dcp1 Dcp2 complexes we see a ~100-fold decrease in KM along with a ~10-fold increase in kmax upon addition of the activator peptide (Supplementary Fig. 5; Supplementary Table 1). Substrate analog positioning suggests a path for RNA binding. This is consistent with recent evidence that Sp Edc1 peptides can Our structure suggests that the first transcribed nucleotide of enhance RNA binding to the decapping complex19,20. RNA binds Phe223 at the top of an extended, positively charged

4 NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x ARTICLE

a c

Dcp2 NRD Dcp2 NRD Dcp1 5’

Edc1 Dcp2 CD Dcp2 CD Lys-rich 90° patch

Edc3 Box B helix Lys-rich patch Edc3 Box B helix 3’

b d

W49 K99 W49 D53 D53 K134 K134 E151 E148

190 s loop 190 s loop 5.4 Å

4.1 Å

E152Q E152Q

F223 R132

Fig. 3 Substrate analog binding to the active conformation of Edc1–Dcp1–Dcp2–Edc3 decapping complex. a Full Edc1–Dcp1–Dcp2–Edc3 structure showing substrate analog binding between the Dcp2 NRD and CD. b Close-up of substrate analog bound at the Dcp2 active site. Conserved, catalytically important residues that contact the cap analog are shown as sticks. The catalytic Glu base (mutated to Gln in our structure) is colored red; Mg2+ ion is shown as a 2+ gray sphere; Fo–Fc omit map for cap analog and Mg is shown at 2.5σ in gray. c Edc1–Dcp1–Dcp2–Edc3 complex with calculated electrostatic surface (−6 to +6 kT/e) in the same orientation as in (a) (left), and rotated 90° (right). Dotted orange line shows the predicted path of RNA based on the orientation of substrate analog in the active site of Dcp2. The 5′ cap of RNA binds in the Dcp2 composite active site (substrate analog in cyan), RNA follows the positively charged Dcp2 CD C-terminal Box B helix past a conserved Lys-rich patch, to Edc3 coactivator. d Alignment of the substrate analog-bound (colored structure, PDB 6AM0) and product-bound (gray, PDB 5LOP 22) K. lactis Dcp2 structures; view is rotated 180° from that shown in (a) and (b). The Dcp2 NRD (residues 8–97) was used for alignment, giving a backbone RMSD of 0.78 Å; all-atom RMSD for alignment of both the NRD and CD (residues 8–240) of substrate analog versus product-bound Dcp2 is 1.6 Å. Distance from catalytic Glu residue to β-phosphate of substrate analog or product are shown as red and gray dotted lines, respectively

C-terminal helix on the CD of Dcp2 (Box B helix, Fig. 3a). Based support to the RNA-binding path that has been hypothesized for on this cap orientation, the RNA body likely extends down the some time by ourselves and others. positively charged Box B helix, past a conserved Lys-rich patch on the dorsal surface of Dcp2, to coactivator Edc3. This predicted Substrate and product-bound Dcp2 conformations are similar. RNA-binding path is consistent with many lines of biophysical Alignment of the Dcp2 catalytic core from the active conforma- and biochemical evidence, including: (1) NMR chemical shift tion substrate analog and product-bound structures shows that perturbations induced by RNA binding on the surface of the these adopt overall very similar Dcp2 domain orientations 19,26 Dcp2 CD , (2) biochemical data showing that the Box B helix (Fig. 3d; Supplementary Fig. 1a). This is consistent with NMR 25 is critical for decapping activity in vitro , (3) kinetic analysis chemical shift data that suggest the Edc1–Dcp1–Dcp2 complex showing that mutations to basic residues in both the Box B helix with either product m7GDP or capped RNA substrate have ﬁ 26 19 and Lys-rich patch signi cantly increase substrate KM , (4) SPR similar conformations in solution . The most notable difference data showing that mutations in the Box B helix signiﬁcantly in the aligned substrate analog and product-bound 17 weaken RNA binding , (5) yeast growth assays showing that Dcp2 structures is the difference in positioning of the phos- mutations in the Box B helix and Lys-rich patch result in phate chain of the cap. In the substrate analog-bound state, the β- 26 temperature-sensitive growth defects , and (6) evidence that phosphate, which undergoes nucleophilic attack by water during Edc3 activates decapping in part by enhancing RNA binding to cap cleavage, is located 5.4 Å from the general base Glu152. This 22 the decapping complex . The positioning of substrate analog in is very similar to the substrate-Glu distance in the prototypical the Dcp2 active site revealed by our Nudix hydrolase MutT, in which a solution-state NMR structure Edc1–Dcp1–Dcp2–Edc3 structure suggests an RNA-binding path showed the catalytic Glu residue to be ~6 Å from the non- along the Dcp2 Box B helix that nicely agrees with all of this hydrolyzable AMPCPP substrate, with an intervening water previous RNA binding and activity data, adding structural molecule well-positioned for nucleophilic attack of the β-

NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x phosphate51. In the post-catalytic, m7GDP product-bound a structure, the β-phosphate of cap is positioned nearer to the Nudix helix, at a distance of 4.1 Å from the general base Glu152. 6 Substrate and product-bound structures also differ in the )

number of magnesium atoms located at the active site. In our –1 substrate analog-bound structure, we were able to build one 4 magnesium atom that is coordinated by Glu148. Although there (min WT + G-RNA – obs is some additional density present in the Fo Fc difference map at k 2 WT + A-RNA the active site (Supplementary Fig. 6a), this was too weak to WT + C-RNA + reliably build additional Mg2 or water. This agrees well with a Y220G + G-RNA 0 Y220G + A-RNA high-resolution structure of the S. cerevisiae Dcp2 CD, which Y220G + C-RNA showed a single magnesium atom coordinated directly to Sc 0.0 0.5 1.0 1.5 2.0 Glu149 (Kl Glu148), which is coordinated by other conserved Glu [Sp Dcp1/Dcp2] (µM) residues through water-mediated contacts. In both the K. lactis and S. pombe product-bound structures, however, three magne- b 8 sium atoms are placed in the Dcp2 active site, coordinated by the conserved CD Glu residues and the phosphate groups of m7GDP (Supplementary Fig. 6b,c). Enzymology from our lab has shown + 6 that the Mg2 -dependence of decapping kinetics ﬁts a hill ) coefﬁcient of 2.4, suggesting Dcp2 binds at least two metal ions52, –1 4 which is most consistent with the product-bound structures. In (min

the end, both our substrate analog-bound structure (2.84 Å max k resolution) and the product-bound structures (3.1 Å and 3.5 Å for 2 PDB 5N2V37 and 5LOP22, respectively) are not at sufﬁcient + resolution to very accurately distinguish Mg2 and water molecules, and high-resolution structures of Dcp2 with bound 0 substrate and product will be required to unambiguously determine the location and precise role of coordinated magnesium and water molecules in the active site.

WT + G-RNAWT + A-RNAWT + C-RNA

Y220G + G-RNAY220G + A-RNAY220G + C-RNA Dcp2 shows selectivity for the first transcribed nucleotide.As discussed above, our structure of substrate analog-bound Fig. 4 Dcp2 selectivity for the first transcribed nucleotide of RNA depends Kl k Dcp2 suggests that the first transcribed nucleotide of RNA is on Tyr220 ( Phe223). a Plots of obs for decapping a 29-mer RNA recognized by the conserved aromatic Phe223 (Sp Tyr220). Sur- substrate with G, A, or C as the first transcribed nucleotide (G-RNA, A- prisingly, although mutation of this residue produces RNA, C-RNA, respectively), with WT or Tyr220Gly S. pombe temperature-sensitive growth defects in yeast26, Tyr220Ala Edc1–Dcp1–Dcp2 complex versus enzyme concentration. The concentration mutations in our single-turnover in vitro decapping assays with S. of the Dcp1–Dcp2(1–243) complex is varied, with Edc1(155–180) µ k pombe Edc1–Dcp1–Dcp2 or Dcp1–Dcp2 complex had no effect coactivator peptide kept at saturating concentration (50 M). Errors on obs k on catalysis (Fig. 2a,b; Supplementary Fig. 4). In the apo, inactive are s.d. of two independent replicates. b Bar graph showing max values k conformation of Dcp1–Dcp2 that predominates in solution19, determined from obs versus enzyme concentration plots in (a). Errors are fi k Tyr220 (Kl Phe223) interacts with conserved aromatic W43 on standard error of the ts to determine max the NRD of Dcp2 (e.g., PDB 2QKM17, 5J3Y20, 5KQ121), which likely stabilizes the inactive conformation of the enzyme. Fur- complexes (Fig. 4; Supplementary Table 2). Tyr220Gly mutations thermore, we have recently shown that Tyr220 plays an impor- were used here to ensure maximal disruption of hydrophobic tant role in mediating Dcp2 conformational dynamics and that stacking interactions with substrate; this mutation also resulted in mutation of this residue alleviates an autoinhibited conformation a small acceleration of decapping, likely resulting from disruption of the apo enzyme42. It is therefore likely that Tyr220 is a multi- of the Y220–W43 interaction in the inactive Dcp2 conformation functional residue, which stabilizes an inactive/autoinhibited discussed above. With WT decapping complex, kmax for both G- conformation of Dcp2 in the apo state and acts as a nucleotide RNA and A-RNA is ~2-fold faster than C-RNA. Upon mutation binding platform in the substrate-bound, active conformation. In of Tyr220, decapping is accelerated and the selectivity of 5′ G/A this way, mutation of Tyr220 may both accelerate catalysis by versus C nucleotides is abolished, resulting in approximately the alleviating autoinhibition and impair catalysis by causing defects same kmax value for G-RNA, A-RNA, and C-RNA. This suggests in substrate binding, resulting in decapping kinetics that appear that Dcp2 discriminates between G/A versus C at the first relatively unaffected by Tyr220Ala mutation. transcribed nucleotide position during the catalytic step of It was recently reported that Dcp2-mediated decapping in decapping, and that this selectivity critically depends on Tyr220 mammals is inhibited by m6A modification of the first (Kl Phe223). Dcp2′s Tyr220-dependent preference for G/A versus transcribed nucleotide of RNA substrates15, suggesting that C at this position is in line with the observation that there is a Dcp2 may be selective for the identity of the first transcribed strong genome-wide bias, in both fission yeast54 and humans55, nucleotide, as is observed in the bacterial decapping enzyme toward purines (G,A) over pyrimidines (C,U) at the transcription NudC53. Accordingly, we next asked if Dcp2 selectivity for the start site +1 position. Furthermore, our selectivity data supports first transcribed nucleotide of RNA substrate depends on Tyr220. the hypothesis that Tyr220 (Kl Phe223) is important for We compared decapping kinetics for cap-radiolabeled 29-mer recognition of the first transcribed nucleotide and that the RNA substrates where the first transcribed nucleotide was either apparent lack of kinetic defects, or even acceleration of guanosine (G-RNA), adenosine (A-RNA), or cytosine (C-RNA) decapping, when this residue is mutated may arise from its using WT, or Tyr220Gly Sp Edc1–Dcp1–Dcp2 decapping

6 NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x ARTICLE multi-functional role in different conformations of the bound PNRC2–Dcp1–Dcp2 structure21, Dcp2 adopts a con- enzyme19,42. formation in which the NRD recognizes one m7G nucleotide of Finally, we also asked whether the activated S. pombe cap analog as in the active conformation, but the CD of Dcp2 is decapping complex showed selectivity for A versus m6A at the rotated 90° and Sp Tyr220 stacks against the opposite face of this first transcribed nucleotide position of RNA, as has been recently m7G, rather than the 190s loop (Fig. 5a, pre-catalytic; Fig. 5c). In shown for mammalian Dcp215. In contrast to the mammalian the pre-catalytic conformation, the second m7G nucleotide of our decapping enzyme, we found that fission yeast Edc1–Dcp1–Dcp2 two-headed cap analog is bound by the strictly conserved Sp complex, or Dcp2 alone, showed no selectivity for A versus m6A Tyr92 residue near the Dcp2 hinge. In the Edc1-bound active at the 5′-most nucleotide position on a 29-mer RNA substrate in conformation, however, this Tyr is occluded by interaction with in vitro decapping assays (Supplementary Fig. 7). This suggests the Dcp2-activating motif on Edc1 (Fig. 5d). It is possible that this that recognition of the m6A modification at the first transcribed pre-catalytic conformation, in which cap analog is bound by a 6 nucleotide of mRNA (m Am in mammals, where this nucleotide composite nucleotide binding site consisting of conserved resi- is also 2′-O-methylated) by Dcp2 may only be conserved in dues on the Dcp2 CD and NRD, represents an early step in cap higher eukaryotes. recognition that precedes formation of the active conformation and its consolidation by the Edc1 YAGxxF motif. Likewise, the apo Edc1–Dcp1–Dcp2 structure by Valkov et al.20, which adopts Discussion a conformation that positions RNA-binding residues on the Dcp2 Several recent studies have provided new structural information CD and Dcp1 opposite each other to create a positively charged – about the mRNA-decapping enzyme Dcp219 22. These include channel, may represent a conformation of the apo enzyme that is structures of the Dcp1–Dcp2 heterodimer with markedly differ- primed for RNA binding or scanning21,24 (Fig. 5a; RNA-binding). ent conformations of Dcp2, bound cap analog or m7GDP pro- With both of these alternative conformations, however, it is so far duct, and in complex with Edc1-like or Edc3 coactivators difficult to rule out the possibility that these are simply off- (Fig. 5a). Here we expand on the structural understanding of pathway conformations captured by crystallization. This is not to Dcp2 activation and catalysis with a new structure of the four- say that valuable information is not provided by conserved cap or protein complex Edc1–Dcp1–Dcp2–Edc3 in the catalytically coactivator contacts with the enzyme, but rather that these spe- active conformation of Dcp2 with a tight-binding substrate ana- cific conformations may not be on-pathway in the Dcp2 catalytic log bound to the enzyme active site (Fig. 1a-c, Table 1). The cycle. Indeed, recent PRE NMR data suggests that the apo Edc1- coactivator Edc1 peptide engages conserved residues on bound conformation in the Valkov et al. structure is not popu- Dcp1–Dcp2 NRD, providing a complementary surface against lated in solution in vitro19. In these same experiments, it was not which the Dcp2 CD and flexible hinge pack in the catalytically possible to distinguish our pre-catalytic substrate-bound con- active conformation, thus consolidating this conformation and formation from the closed, inactive conformation of Dcp2 that promoting catalysis (Fig. 1d,e and 2) Cap substrate analog binds predominates in solution (Fig. 5a, inactive; Fig. 5b). to the enzyme active site in an orientation that positions the RNA Building on recent reports of Dcp2 autoinhibition in yeast29, body to follow an extended, positively charged helix of Dcp2 to our lab has now shown that the C-terminal extension in the bound Edc3 coactivator (Fig. 3), which has been shown to fission yeast Dcp2 contains autoinhibitory motifs that interact enhance RNA binding to these complexes in vitro22. Finally, we with the core domains of Dcp1–Dcp2 to inhibit the decapping show that the conserved aromatic residue Tyr220 (Kl Phe223) complex and impair activation by Edc1, and that this that binds the second nucleotide of our substrate analog is critical autoinhibition is alleviated by coactivator Edc342. The for recognition of the first transcribed nucleotide in RNA by the Edc1–Dcp1–Dcp2–Edc3 structure presented here provides the decapping complex (Fig. 4). Purine specificity is conferred for this structural framework to begin to understand how coactivators act nucleotide during the catalytic step of decapping. to alleviate autoinhibition and activate catalysis. We suggest a Many different conformations of the decapping enzyme Dcp2 model of decapping regulation whereby full-length Dcp2 is have been structurally characterized, but a consensus is now autoinhibited and coactivators Edc3 and Edc1 act simultaneously emerging around the identity of the catalytically active con- to (1) alleviate this autoinhibition, and (2) activate decapping formation of Dcp2. With the addition of the catalysis by both consolidating the Dcp2 active conformation and Edc1–Dcp1–Dcp2–Edc3 cap analog structure presented here, we by enhancing RNA binding to the complex. In this way, switch- now have both product m7GDP and substrate analog-bound like mutli-log unit changes in decapping activity are easily structures of the Dcp1–Dcp2 decapping complex in what can achievable when multiple coactivators bind Dcp2 and function safely be described as the active conformation (Fig.5a, active simultaneously. Because many decapping coactivators have – structures). The substrate and product-bound active conforma- genetic or physical interactions with one another29,45 49,we tions are nearly identical, with major differences only in the anticipate that combinatorial control of decapping via multiple positioning of the cap phosphate chain before and after hydrolysis coactivators is a general and important principle for regulation of (Fig. 3d). These structures all show that the m7G nucleotide of mRNA decay. cap is specifically recognized by stacking with a conserved Trp and hydrogen bonding to a conserved Asp residue on the Dcp2 Methods NRD, explaining why the NRD contributes 100-fold to cata- Protein expression and purification. K. lactis his-tev-Dcp1(1–188)–Dcp2(1–275, 31 lysis . In the active conformation, the 190s loop of the Dcp2 CD E152Q)-Edc3(1–66) polycistronic construct used for crystallization experiments packs against the opposite face of the m7G base, which brings the was obtained as a single E. coli codon-optimized DNA sequence with a T7 pro- Nudix helix and catalytic Glu within ~6 Å of the cap β-phosphate, moter preceding each protein as a gBlock from Integrated DNA Technologies (his is hexahistadine affinity tag, tev is Tobacco Etch Virus protease cleavage site; see close enough to carry out hydrolysis chemistry and cap cleavage Supplementary Table 3 for sequence). This was cloned into a Kanamycin-resistant (Fig. 3b). expression vector and transformed into E. coli BL21-star DE3 cells (Invitrogen; see The catalytic importance and roles of the additional con- Supplementary Table 4 for primers used). Cells were grown in LB media to OD formations of Dcp2 that have been structurally characterized, but ~0.6, protein expression was induced with IPTG for 18 h at 20 °C. Cells were harvested at 5000 × g, lysed by sonication, and clarified at 16,000 × g in lysis buffer are very different from the active conformation (Fig. 5a, inactive, (50 mM sodium phosphate pH 7.5, 300 mM sodium chloride, 10 mM imidazole, RNA-binding, and pre-catalytic structures; Fig. 5b,c), remain less 5% glycerol, 10 mM 2-mercaptoethanol, Roche EDTA-free protease inhibitor clear. In our previously published pre-catalytic substrate analog- cocktail). The protein complex was purified by Ni-NTA affinity chromatography

NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x

Nucleotides ATP Cap substrate analog m7GDP product m7GDP product Cap substrate analog Coactivators Edc1 Edc1 Edc1 Edc3 Edc3 Conformation Inactive RNA-binding? Pre-catalytic? Active Active Active Organism S. pombe S. pombe S. pombe K. lactis S. pombe K. lactis PDB ID 2QKM 5J3T 5KQ4 5LOP 5N2V 6AM0 Citation She et al. Valkov et al. Mugridge et al. Charenton et al. Wurm et al. Mugridge et al. 2008 2016 2016 2016 2017 2018 bcd

W43

W43 Y220 Y92 Y92 Y220

Y220

Fig. 5 Summary of recent Dcp1–Dcp2 structures and interactions between important aromatic cap-binding residues in different conformations. a Gallery showing the different overall conformations of recent Dcp1–Dcp2 structures. Text lists bound nucleotides, coactivators, and proposed description of the Dcp2 conformation. Dcp1 is yellow, Dcp2 NRD is purple, Dcp2 CD is green with the catalytic Nudix helix colored red, Edc1 is brown, Edc3 is dark blue, cap substrate analog is cyan, m7GDP product is pink. From left to right, PDB codes and citations are 2QKM17, 5J3T20, 5KQ421, 5LOP22, 5N2V19, 6AM0 (this work). b Close-up view of ATP-bound inactive structure (PDB 2QKM)17. Conserved, cap-binding residues Trp43 and Tyr220 make edge-on contacts in this inactive conformation. ATP is gray. c Close-up view of cap analog-bound, pre-catalytic structure (PDB 5KQ4)21. Residues Tyr220 and Trp43 bind one m7G nucleotide of two-headed cap analog, Tyr92 binds the second m7G nucleotide of the cap analog. d Close-up view of cap analog-bound, active structure (PDB 6AM0); S. pombe residue numbering is used here instead of K. lactis for comparison with structures shown in (b) and (c). Trp43 (Kl Trp49) binds one m7G nucleotide of two-headed cap analog and Tyr220 (Kl Phe223) binds the second m7G of cap analog, mimicking binding of the ﬁrst transcribed nucleotide. Tyr92 (Kl Tyr98) is buried in a protein–protein interface with Edc1 coactivator in the active conformation

fi and solubility/affinity tags were cleaved by treatment with TEV protease overnight sequencing, and puri ed as above (see Supplementary Table 4 for primer at room temperature in elution buffer (50 mM sodium phosphate pH 7.5, 300 mM sequences). sodium chloride, 250 mM imidazole, 10 mM 2-mercaptoethanol). The Kl Dcp1–Dcp2–Edc3 complex was separated from cleaved tags by size exclusion chromatography on a GE Superdex 200 16/60 gel filtration column in crystal- Crystallization of Edc1–Dcp1–Dcp2–Edc3 with substrate analog. Kl Edc1(355- lization buffer (10 mM HEPES pH 7, 150 mM NaCl, 1 mM DTT). The purified 380) peptide (from gene KLLA0_A01474g) was chemically synthesized by Peptide complex was concentrated to 15 mg/mL and flash frozen in liquid nitrogen for 2.0 and dissolved in crystallization buffer to a concentration of 5 mM. Two-headed crystallization experiments. cap substrate analog, m7Gp(S)ppp(S)m7G50,56, was dissolved in water at a con- S. pombe his-GB1-tev–Dcp1(1–127)–Dcp2(1–243) constructs used for kinetic centration of 100 mM and the pH was adjusted to ~7. experiments were obtained by cloning the genomic Sp Dcp2(1–243) coding Protein solution consisting of 4 mg/mL purified K. lactis Dcp1(1–188)–Dcp2 sequence into MCS2 of Novagen pACYC co-expression vector, and an E. Coli (1–275, E152Q)–Edc3(1–66), 1 mM Kl Edc1 peptide, 6 mM substrate analog, and codon-optimized Sp Dcp1(1–127) DNA sequence obtained from Integrated DNA 10 mM MgCl was prepared in crystallization buffer and incubated at room Technologies into MCS1 of the pACYC vector containing an N-terminal his-GB1- 2 temperature for 30 min. Using a TTP Labtech Mosquito Crystal robot, 250 nL of tev tag (GB1 is a solubility tag derived from the B1 domain from Protein G; see protein solution was combined with 250 nL of well solution, and then 50 nL of seed Supplementary Table 3 for codon-optimized Sp Dcp1 sequence). The Sp stock, in a hanging-drop vapor diffusion experiment. Crystallization drops were Dcp1–Dcp2 protein complex was expressed and purified as described above for the prepared at room temperature and immediately moved to 4 °C. The well solution Kl Dcp1–Dcp2–Edc3 complex. After TEV cleavage, the complex was separated contained 220 mM magnesium acetate, 30 mM EDTA, and 8% w/v PEG 8000. The from cleaved tags by size exclusion chromatography on a GE Superdex 200 16/60 seed stock was prepared using Hampton’s Seed Bead kit, using crystals grown with gel filtration column in storage buffer (50 mM HEPES pH 7.5, 100 mM NaCl, 5 1:1 protein solution and well solution containing 200 mM magnesium acetate and mM DTT). The purified Sp Dcp1–Dcp2 complex was concentrated, glycerol added 10% w/v PEG 8000 at room temperature, diluted 10,000-fold. Block-shaped crystals to a final concentration of 20%, and flash frozen in liquid nitrogen; final ~40–80 µm in size grew at 4 °C within 1–2 days. Crystals were flash frozen in liquid concentration was 200 µM. Point mutants of the Sp Dcp1–Dcp2 complex were nitrogen using a cryoprotectant consisting of well solution with 15% PEG 8000. made by whole plasmid PCR with mutagenic divergent primers, verified by

8 NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x ARTICLE

Crystallographic data collection and refinement. All data were collected at Received: 25 August 2017 Accepted: 22 February 2018 Lawrence Berkeley National Lab, Beamline 8.3.1 at the Advanced Light Source on a Dectris Pilatus3 S 6 M detector at 100 K and a wavelength of 1.11583 Å. Data were indexed, integrated and scaled using XDS57. The structures were solved in space group C2 by molecular replacement using Phaser in the CCP4 suite58. Fragments of chains A, B, and C of PDB 5LOP were used as molecular replacement models22, fi – – – rst placing Dcp1 Dcp2 NRD(1 96), then Dcp2 CD(104 240), then Dcp2 References (252–260)–Edc3(1–66). The asymmetric unit contained two nearly symmetric – – – 1. Moore, M. J. From birth to death: the complex lives of eukaryotic mRNAs. copies of the Edc1 Dcp1 Dcp2 Edc3 complex. In the crystal lattice, one molecule – of Edc3 Lsm domain forms a dimer as in the Kl Edc3–Dcp1–Dcp2–m7GDP Science 309, 1514 1518 (2005). 22 2. Parker, R. RNA degradation in Saccharomyces cerevisae. Genetics 191, product structure (PDB 5LOP) , but the spacegroups (C2 for AM60 and P6422 for – 5LOP) and overall crystal packing are entirely different between our structure and 671 702 (2012). the previously determined K. lactis structure. Molecular replacement failed to 3. Dunckley, T. & Parker, R. The DCP2 protein is required for mRNA decapping correctly place the Dcp2(252–260)–Edc3(1–66) fragment for chain H, as the in Saccharomyces cerevisiae and contains a functional MutT motif. EMBO J. – density was very weak for this chain of Edc3, likely because this molecule of Edc3 18, 5411 5422 (1999). does not form a dimer in the crystal lattice. This fragment was placed manually in 4. Arribas-Layton, M., Wu, D., Lykke-Andersen, J. & Song, H. Structural and COOT, based on strong density for the Dcp2(252–260) Edc3-binding motif and functional control of the eukaryotic mRNA decapping machinery. Biochim. strong density for some β-strands in Edc3. After a preliminary round of refinement Biophys. Acta 580–589, 2013 (1829). including rigid-body refinement in PHENIX59, the Edc1 peptide, Dcp2 hinge, and 5. Li, Y. & Kiledjian, M. Regulation of mRNA decapping. Wiley Interdiscip. Rev. Dcp2 extended helix were built manually in COOT60. The structure was then 1, 253–265 (2010). iteratively refined in PHENIX and manually adjusted in COOT. Final refinement 6. Nagarajan, V. K., Jones, C. I., Newbury, S. F. & Green, P. J. XRN 5′ → 3′ used Non-Crystallographic Symmetry, Translation-Libration-Screw parameteriza- exoribonucleases: structure, mechanisms and functions. Biochim. Biophys. tion for chain H, and secondary structure, stereochemical and atomic displacement Acta 1829, 590–603 (2013). parameter restraints. For the final model of Kl Edc1–Dcp1–Dcp2–Edc3, all residues 7. Kervestin, S. & Jacobson, A. NMD: a multifaceted response to premature were in favored (95.6%) or allowed (4.4%) regions of the Ramachandran plot as translational termination. Nat. Rev. Mol. Cell Biol. 13, 700–712 (2012). evaluated by MolProbity61. 8. Popp, M. W.-L. & Maquat, L. E. Organizing principles of mammalian nonsense-mediated mRNA decay. Annu. Rev. Genet. 47, 139–165 (2013). 9. Cho, H., Kim, K. M. & Kim, Y. K. Human proline-rich nuclear receptor Structural analysis and visualization fi . All alignments and structural gures were coregulatory protein 2 mediates an interaction between mRNA surveillance prepared using PyMol version 1.7. Electrostatic surfaces were calculated using machinery and decapping complex. Mol. Cell 33,75–86 (2009). PDB2PQR & APBS webservers (http://nbcr-222.ucsd.edu/pdb2pqr_2.0.0/)62,63, 10. Karousis, E. D., Nasif, S. & Mühlemann, O. Nonsense-mediated mRNA decay: using the AMBER94 forcefield to calculate charges. Electrostatic surfaces were novel mechanistic insights and biological impact. Wiley Interdiscip. Rev. RNA visualized in PyMol using APBS tools 2.1. Buried surface area in protein interfaces 7, 661–682 (2016). was calculated using the PDBePISA webserver (http://www.ebi.ac.uk/pdbe/ 11. Guo, H., Ingolia, N. T., Weissman, J. S. & Bartel, D. P. Mammalian prot_int/pistart.html)64. microRNAs predominantly act to decrease target mRNA levels. Nature 466, 835–840 (2010).

Kinetic decapping assays. For the decapping assays measuring kobs for 12. Jonas, S. & Izaurralde, E. Towards a molecular understanding of microRNA- Dcp1–Dcp2 point mutants ±Edc1 described in Fig. 2 and Supplementary Fig. 4, Sp mediated gene silencing. Nat. Rev. Genet. 16, 421–433 (2015). Dcp1–Dcp2(1–243) WT and mutant complexes had a final concentration of 2 µM 13. Geisler, S., Lojek, L., Khalil, A. M., Baker, K. E. & Coller, J. Decapping of long and Sp Edc1 (155–180) peptide had a final concentration of 100 µM. Sp Edc1 non-coding RNAs regulates inducible genes. Mol. Cell 45, 279–291 (2012). (155–180) peptide was chemically synthesized by Peptide 2.0 and stock solutions 14. Presnyak, V. et al. Codon optimality is a major determinant of mRNA were prepared at 5 mM concentration in crystallization buffer, flash frozen in liquid stability. Cell 160, 1111–1124 (2015). nitrogen, and stored at −20 °C. These experiments used an in vitro transcribed, 15. Mauer, J. et al. Reversible methylation of m6Am in the 5′ cap controls mRNA enzymatically capped 355-mer RNA substrate derived from MFA2 and 32P-labeled stability. Nature 541, 371–375 (2017). at the α-phosphate of cap. Reactions were carried out in the decapping buffer used 16. She, M. et al. Crystal structure and functional analysis of Dcp2p from previously to measure PNRC2 activation kinetics (50 mM HEPES pH 7, 100 mM Schizosaccharomyces pombe. Nat. Struct. Mol. Biol. 13,63–70 (2006). KCl, 10% glycerol, 5 mM MgCl2, with RNase inhibitor and 0.1 mg/mL acetylated 17. She, M. et al. Structural basis of Dcp2 recognition and activation by Dcp1. 21 BSA) . 3× protein solutions with Dcp1–Dcp2+Edc1 peptide, and 1.5× capped Mol. Cell 29, 337–349 (2008). RNA solutions were equilibrated separately at 4 °C for 30 min prior to initiation. 18. Floor, S. N., Borja, M. S. & Gross, J. D. Interdomain dynamics and Reactions were initiated by combining 15 µL 3× protein solution with 30 µL 1.5× coactivation of the mRNA decapping enzyme Dcp2 are mediated by a RNA solution at 4 °C. Time points were quenched by the addition of excess EDTA, gatekeeper tryptophan. Proc. Natl. Acad. Sci. USA 109, 2872–2877 (2012). 7 TLC was used to separate RNA from product m GDP, and the decapped fraction 19. Wurm, J. P., Holdermann, I., Overbeck, J. H., Mayer, P. H. O. & Sprangers, R. fi was quanti ed using a GE Typhoon scanner and ImageQuant software. Plots of Changes in conformational equilibria regulate the activity of the Dcp2 7 fi fi fraction m GDP product versus time were ttoa rst order exponential to obtain decapping enzyme. Proc. Natl. Acad. Sci. USA 114, 6034–6039 (2017). kobs; in the case of R33A and I96G point mutants, decapping kinetics were too slow fi fi 20. Valkov, E. et al. Structure of the Dcp2-Dcp1 mRNA-decapping complex in the to obtain reliable exponential ts and kobs was obtained from a linear t of initial activated conformation. Nat. Struct. Mol. Biol. 23, 574–579 (2016). rates (less than 20% decapped), by dividing the slope by the average endpoint of 21. Mugridge, J. S., Ziemniak, M., Jemielity, J. & Gross, J. D. Structural basis of 0.95. Errors on individual time points in Fig. 2a are the s.d. between two inde- mRNA-cap recognition by Dcp1-Dcp2. Nat. Struct. Mol. Biol. 23, 987–994 pendent repetitions of the decapping reaction; errors on k in Fig. 2b are s.d. of obs (2016). k values obtained from these two independent replicates. obs 22. Charenton, C. et al. Structure of the active form of Dcp1-Dcp2 decapping For decapping assays measuring kinetics with G versus C at the first transcribed enzyme bound to m7GDP and its Edc3 activator. Nat. Struct. Mol. Biol. 23, nucleotide described in Fig. 4 and those measuring KM and kmax for Edc1 activation – (Supplementary Fig. 5, Supplementary Table 1), synthetic 5′-triphosphate 29-mer 982 986 (2016). RNA oligonucleotides with either guanosine (G-RNA) or cytosine (C-RNA) as the 23. Coller, J. mRNA decapping in 3D. Nat. Struct. Mol. Biol. 23, 954 (2016). 5′ nucleotide were obtained from Trilink BioTechnologies and enzymatically 24. Valkov, E., Jonas, S. & Weichenrieder, O. Mille viae in eukaryotic mRNA – capped with 32P-radiolabeled GTP, as above. In the decapping reaction, Sp decapping. Curr. Opin. Struct. Biol. 47,40 51 (2017). Dcp1–Dcp2(1–243) had a final concentration ranging from 2 µM to 4 nM, the final 25. Piccirillo, C., Khanna, R. & Kiledjian, M. Functional characterization of the – concentration of Sp Edc1(155–180) peptide was held constant at 50 µM, and mammalian mRNA decapping enzyme hDcp2. RNA 9, 1138 1147 (2003). capped RNA concentration was <100 pM. For these experiments, glycerol was 26. Deshmukh, M. V. et al. mRNA decapping is promoted by an RNA-binding removed from the decapping buffer described above in order to obtain more easily channel in Dcp2. Mol. Cell 29, 324–336 (2008). saturable kinetics with the 29-mer RNA substrate. Decapping reactions were 27. Aglietti, R. A., Floor, S. N., McClendon, C. L., Jacobson, M. P. & Gross, J. D. prepared, executed, and analyzed as described above. To obtain kmax and KM, kobs Active site conformational dynamics are coupled to catalysis in the mRNA fi = + – was plotted versus protein concentration and t to the model: kobs kmax[E]/KM decapping enzyme Dcp2. Structure 21, 1571 1580 (2013). [E]. Errors on kobs values shown in Fig. 4a are s.d. of two independent replicates. 28. Mildvan, A. S. et al. Structures and mechanisms of Nudix hydrolases. Arch. fi – Errors on kmax values shown in Fig. 4b are the standard error from ts to determine Biochem. Biophys. 433, 129 143 (2005). kmax and KM. 29. He, F. & Jacobson, A. Control of mRNA decapping by positive and negative regulatory elements in the Dcp2 C-terminal domain. RNA 21, 1633–1647 (2015). Data availability. Coordinates and structure factors were deposited in the Protein 30. Fromm, S. A. et al. The structural basis of Edc3‐ and Scd6‐mediated activation Data Bank with accession code PDB 6AM0. Other data are available from the of the Dcp1:Dcp2 mRNA decapping complex. EMBO J. 31, 279–290 (2012). corresponding author upon reasonable request.

NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03536-x

31. Floor, S. N., Jones, B. N., Hernandez, G. A. & Gross, J. D. A split active site 59. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for couples cap recognition by Dcp2 to activation. Nat. Struct. Mol. Biol. 17, macromolecular structure solution. Acta Crystallogr. D Biol. Crystallogr. 66, 1096–1101 (2010). 213–221 (2010). 32. Chang, C.-T., Bercovich, N., Loh, B., Jonas, S. & Izaurralde, E. The activation 60. Emsley, P., Lohkamp, B., Scott, W. G. & Cowtan, K. Features and development of the decapping enzyme DCP2 by DCP1 occurs on the EDC4 scaffold and of Coot. Acta Crystallogr. D Biol. Crystallogr. 66, 486–501 (2010). involves a conserved loop in DCP1. Nucleic Acids Res. 42, 5217–5233 (2014). 61. Chen, V. B. et al. MolProbity: all-atom structure validation for 33. Jonas, S. & Izaurralde, E. The role of disordered protein regions in the macromolecular crystallography. Acta Crystallogr. D Biol. Crystallogr. 66, assembly of decapping complexes and RNP granules. Genes Dev. 27, 12–21 (2010). 2628–2641 (2013). 62. Dolinsky, T. J. et al. PDB2PQR: expanding and upgrading automated 34. Dunckley, T., Tucker, M. & Parker, R. Two related proteins, Edc1p and Edc2p, preparation of biomolecular structures for molecular simulations. Nucleic stimulate mRNA decapping in Saccharomyces cerevisiae. Genetics 157,27–37 Acids Res. 35, W522–W525 (2007). (2001). 63. Baker, N. A., Sept, D., Joseph, S., Holst, M. J. & McCammon, J. A. 35. Borja, M. S., Piotukh, K., Freund, C. & Gross, J. D. Dcp1 links coactivators of Electrostatics of nanosystems: application to microtubules and the ribosome. mRNA decapping to Dcp2 by proline recognition. RNA 17, 278–290 (2011). Proc. Natl. Acad. Sci. USA 98, 10037–10041 (2001). 36. Lai, T. et al. Structural basis of the PNRC2-mediated link between mRNA 64. Krissinel, E. & Henrick, K. Inference of macromolecular assemblies from surveillance and decapping. Structure 20, 2025–2037 (2012). crystalline state. J. Mol. Biol. 372, 774–797 (2007). 37. Wurm, J. P., Overbeck, J. & Sprangers, R. The S. pombe mRNA decapping complex recruits cofactors and an Edc1-like activator through a single dynamic surface. RNA 22, 1360–1372 (2016). Acknowledgements 38. Kshirsagar, M. & Parker, R. Identification of Edc3p as an enhancer of mRNA We thank J. Holton and G. Meigs at Lawrence Berkeley National Laboratory, Advanced decapping in Saccharomyces cerevisiae. Genetics 166, 729–739 (2004). Light Source beamline 8.3.1, and X. Liu for help with X-ray data collection and refine- 39. Badis, G., Saveanu, C., Fromont-Racine, M. & Jacquier, A. Targeted mRNA ment. We also thank J. Kowalska for helpful discussions on the design and synthesis of degradation by deadenylation-independent decapping. Mol. Cell 15,5–15 the two-headed cap analog. This work was supported by the US National Institutes of (2004). Health (R01 GM078360 to J.D.G. and NRSA fellowship F32 GM105313 to J.S.M.), The 40. Dong, S. et al. YRA1 autoregulation requires nuclear export and cytoplasmic Foundation for Polish Science (TEAM/2016-2/13 to J.J.), the National Science Centre, Edc3p-mediated degradation of its pre-mRNA. Mol. Cell 25, 559–573 (2007). Poland (fellowship no. UMO-2014/12/T/NZ1/00528 to M.Z.), the Genentech Foundation 41. Harigaya, Y., Jones, B. N., Muhlrad, D., Gross, J. D. & Parker, R. Identification Fellowship (to R.W.T), and the UCSF Discovery Fellowship (to R.W.T). The Advanced and analysis of the interaction between Edc3 and Dcp2 in Saccharomyces Light Source is supported by the US Department of Energy under contract no. DE-AC02- cerevisiae. Mol. Cell. Biol. 30, 1446–1456 (2010). 05CH11231. 42. Paquette, D. R., Tibble, R. W., Daifuku, T. S. & Gross, J. D. Control of mRNA decapping by autoinhibition. Preprint at https://doi.org/10.1101/266718 Author contributions (2018). J.S.M. designed and purified all protein constructs, carried out crystallization experi- 43. Fenger-Grøn, M., Fillman, C., Norrild, B. & Lykke-Andersen, J. Multiple ments, collected and refined crystallographic data, wrote the manuscript, and prepared processing body factors and the ARE binding protein TTP activate mRNA the figures. J.S.M. and R.W.T. carried out decapping kinetics experiments. M.Z. and J.J. decapping. Mol. Cell 20, 905–915 (2005). designed and synthesized the two-headed cap analog. J.D.G. supervised the project and 44. Yu, J. H., Yang, W.-H., Gulick, T., Bloch, K. D. & Bloch, D. B. Ge-1 is a central experimental design and guided manuscript preparation and editing. All authors read component of the mammalian cytoplasmic mRNA processing body. RNA 11, and commented on the manuscript. 1795–1802 (2005). 45. Eulalio, A. et al. Target-specific requirements for enhancers of decapping in miRNA-mediated gene silencing. Genes Dev. 21, 2558–2570 (2007). Additional information 46. Gavin, A.-C. et al. Functional organization of the yeast proteome by systematic Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- analysis of protein complexes. Nature 415, 141–147 (2002). 018-03536-x. 47. Krogan, N. J. et al. Global landscape of protein complexes in the yeast Saccharomyces cerevisiae. Nature 440, 637–643 (2006). Competing interests: The authors declare no competing interests. 48. Collins, S. R. et al. Functional dissection of protein complexes involved in yeast chromosome biology using a genetic interaction map. Nature 446, Reprints and permission information is available online at http://npg.nature.com/ 806–810 (2007). reprintsandpermissions/ 49. Decourty, L. et al. Linking functionally related genes by sensitive and quantitative characterization of genetic interaction profiles. Proc. Natl. Acad. Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in Sci. USA 105, 5821–5826 (2008). published maps and institutional affiliations. 50. Ziemniak, M. et al. Two-headed tetraphosphate cap analogs are inhibitors of the Dcp1/2 RNA decapping complex. RNA 22, 518–529 (2016). 51. Lin, J., Abeygunawardana, C., Frick, D. N., Bessman, M. J. & Mildvan, A. S. Solution structure of the quaternary MutT−M2+ −AMPCPP−M2+ complex Open Access This article is licensed under a Creative Commons and mechanism of its pyrophosphohydrolase action. Biochemistry 36, Attribution 4.0 International License, which permits use, sharing, 1199–1211 (1997). adaptation, distribution and reproduction in any medium or format, as long as you give 52. Jones, B. N., Quang-Dang, D.-U., Oku, Y. & Gross, J. D. in Methods in appropriate credit to the original author(s) and the source, provide a link to the Creative Enzymology (eds Maquat, L.E. and Kiledjian, M.) Vol. 448, 23–40 (Academic Commons license, and indicate if changes were made. The images or other third party Press, Cambridge, Massachusetts, 2008). material in this article are included in the article’s Creative Commons license, unless 53. Höfer, K. et al. Structure and function of the bacterial decapping enzyme indicated otherwise in a credit line to the material. If material is not included in the NudC. Nat. Chem. Biol. 12, 730–734 (2016). article’s Creative Commons license and your intended use is not permitted by statutory 54. Li, H. et al. Genome-wide analysis of core promoter structures in regulation or exceeds the permitted use, you will need to obtain permission directly from – Schizosaccharomyces pombe with DeepCAGE. RNA Biol. 12, 525 537 (2015). the copyright holder. To view a copy of this license, visit http://creativecommons.org/ 55. Carninci, P. et al. Genome-wide analysis of mammalian promoter architecture licenses/by/4.0/. and evolution. Nat. Genet. 38, 626–635 (2006). 56. Ziemniak, M. et al. Phosphate-modified analogues of m(7)GTP and m(7) Gppppm(7)G-synthesis and biochemical properties. Bioorg. Med. Chem. 23, © The Author(s) 2018 5369–5381 (2015). 57. Kabsch, W. XDS. Acta Crystallogr. D Biol. Crystallogr. 66, 125–132 (2010). 58. McCoy, A. J. et al. Phaser crystallographic software. J. Appl. Crystallogr. 40, 658–674 (2007).

10 NATURE COMMUNICATIONS | (2018) 9:1152 | DOI: 10.1038/s41467-018-03536-x | www.nature.com/naturecommunications